User Modeling Using Conceptual Graphs for Intelligent Agents

We describe a novel inference method to help us deal with the problem of having sparse information about the user. The information we will have available for the user will form his/her model. That model will be represented in a conceptual graph format. We will also have gathered information about categories of users that share characteristics, preferences and interests. These will form the ”prototypes”, as we call them and will be represented in graphs as well. Because we will know significantly more about the prototypes, they will be the source where we will try to get the information we want for the user. This method has been implemented in FRIL [2, 3, 10].